[m-rev.] diff: solver type cast fns for non-C backends / fix test failure

Julien Fischer juliensf at csse.unimelb.edu.au
Mon Jul 4 12:55:04 AEST 2011


On Mon, 4 Jul 2011, Julien Fischer wrote:

>
> Branches: main, 11.07
>
> Fix the failure of hard_coded/solver_default_eq_cmp in deep profiling grades.
> The problem was the test didn't fit into any of the existing categories in 
> the
> Mmakefile in tests/hard_coded -- specifically, we don't have a separate
> category for non-deep profiling, C backend-only test; which was the set of
> grades in which this test would work.
>
> The fix is to remove the C backend-only dependency for this test by 
> generating
> solver type cast functions for the non-C backends and the shifting the test
> into the existing set of tests that should only be run in non-deep profiling
> grades.
>
> compiler/add_solver.m:
> 	Generate solver type cast functions in non-C grades.
>
> tests/hard_coded/Mmakefile:
> 	Don't run solver_default_eq_comp in deep profiling grades.
>
> 	Put a list of tests back into order.

Actually, this doesn't fix the test case failure, since solver types
cause additional problems further along in the MLDS code generator.

I'll commit the changes to add_solver and add a separate workaround for
the test case failure.

Julien.
--------------------------------------------------------------------------
mercury-reviews mailing list
Post messages to:       mercury-reviews at csse.unimelb.edu.au
Administrative Queries: owner-mercury-reviews at csse.unimelb.edu.au
Subscriptions:          mercury-reviews-request at csse.unimelb.edu.au
--------------------------------------------------------------------------



More information about the reviews mailing list